Power curves comparing different global testing methods for detecting a trans-eQTL.

Each panel shows power for detecting a trans-eQTL plotted against the significance threshold of the association test, for 6 of the 8 methods considered. In order to make the plots less cluttered, the power of the Simes method was not plotted, as it was approximately equal to that of Min-P in all cases, and similarly, the power of the ARCHIE method was not plotted, as it was approximately equal to that of Sum-χ² in all cases (see Tables A—F of S1 Text). For ADELLE, q = 0.2 is used. In each panel, power is based on 10³ simulated replicates. Each panel shows the plot for a setting in which a given number of expression traits are associated with the tested trans-eQTL. For each point of the plot, the corresponding vertical bar represents the 95% confidence interval for power. In Panel A, the number of associated expression traits is 5. In Panels B, C, D, E, and F, the numbers of associated expression traits are 10, 20, 50, 100, and 200, respectively.